Tzafrir, your assumptions would be true if the guest doing the "bench mark"
were the only one in the box. However, the benchmark appears to be run in a
resource constrained environment with the benchmark guest competing with
other guests.
The measurements I ran of the benchmark as a VM EC (and an LPAR), with no
resource contention, indicated that the benchmark showed little deviation
from one run to the next, so I would conclude that in the original case
there is other interferance outside the guest that is causing the
inconsistent results. The most obvious is the over commitment of memory.
The benchmark perl progam may appear to be small, but the perl and linux
underneath to make it run that is may not be.
The load numbers stated were probably averages over a period of time, not
peak usage values. If the benchmark were run with other activity that may
or may not be peaking, I would expect inconsistent results.
In an LPAR, with shared CP, other LPARs could cause inconsistent results.
The only way to ensure consistent results is in an LPAR, native or under
VM, with dedicated CP and no other activity from competing guests. (I.e.,
performance measurement 101).
Regards, Jim
Linux S/390-zSeries Support, SEEL, IBM Silicon Valley Labs
t/l 543-4021, 408-463-4021, [EMAIL PROTECTED]
*** Grace Happens ***
Tzafrir Cohen
<[EMAIL PROTECTED] To: [EMAIL PROTECTED]
.ac.il> cc:
Sent by: Linux on Subject: Re: offloading CPU intensive
loads from zLinux to cheaper pastures
390 Port
<[EMAIL PROTECTED]
IST.EDU>
06/20/2003 08:13
AM
Please respond to
Linux on 390 Port
On Thu, Jun 19, 2003 at 09:40:49AM -0700, Jim Sibley wrote:
> With 1 Gig of real memory and 512 MB per guest, you're probably measuring
> the VM paging subsystem or some other overhead phenonmena, which is
> probably tunable, not the Linux guest. - with 10x512 MB guest to 1 GB
real
> memory, you may be overcommited by more than 5 to 1 because VM chews up a
> good bit of that memory too! The miracle would probaly be to do something
> like add more memory.
Sorry, but I fail to understand this epxlanation.
Does the VM always swap-out complete guests?
Because in this case the code and data involved in the calculation (for
some 14 minutes or so) were located in a very small number of pages.
--
Tzafrir Cohen +---------------------------+
http://www.technion.ac.il/~tzafrir/ |vim is a mutt's best friend|
mailto:[EMAIL PROTECTED] +---------------------------+
